Introduction

Irinotecan, also known as CPT-11, is a semi-synthetic analog of camptothecin that is a natural alkaloid extract of plants such as Camtotheca acuminate.1 As an anti-neoplastic agent of the class of topoisomerase 1 inhibitor, irinotecan has been widely used for the treatment of colorectal cancer, non-small-cell lung cancer and several other solid tumors.2, 3 However, the vast interindividual variability in the pharmacokinetics and pharmacodynamics of irinotecan and its metabolites rendered currently used irinotecan dosing strategy, which is solely based on the body surface area of patients, insufficient to perform personalized irinotecan therapy.4, 5 Consequently, the use of irinotecan is often limited by the unpredictable dose-limiting and life-threatening toxicities such as diarrhea6 and myelosuppression that manifest as severe neutropenia, leukopenia, anemia and thrombocytopenia.7

Recent studies have implied not only that the increased systemic level of SN-38, the active metabolite of irinotecan, is correlated with irinotecan-induced severe myelosuppression8 but also that the increased local level of SN-38 in the intestine was associated with life-threatening delayed-onset diarrhea.9, 10 Therefore, genetic polymorphisms in genes involved in metabolizing and transporting irinotecan, which may alter the pharmacokinetics of SN-38, have been extensively studied for their association with the interindividual variability in clinical outcome and toxicity of irinotecan-based therapy in recent years.5, 11, 12

The proteins that have critical roles in irinotecan metabolism include carboxylesterases (CES) and the two members of the cytochrome P450 family, CYP3A4 and CYP3A5. Although CES bio-transforms irinotecan into SN-38 that is 100- to 1000-fold more potent than irinotecan,13 CYP3A4 and CYP3A5 oxidize irinotecan to form two other metabolites, 7-ethyl-10-[4-N-(5-aminopentanoic acid)-1-piperidino]-carbonyloxycamptothecin and 7-ethyl-10-[4-(1-piperidino)-1-amino]-carbonyloxycamptothecin (NPC) that lack anticancer activity. However, it is now known that NPC can also be converted to SN-38 by CES.14, 15

In addition, genes such as UGT1A1, UGT1A7, UGT1A9, ABCC2, ABCG2 and ABCB1 encode proteins that have important roles in the detoxification and hepatobiliary disposition of irinotecan. Glucuronidation of SN-38 to form the inactive SN-38G by uridine diphosphate glucuronosyltransferase isoforms, primarily by UGT1A1 in the liver, is the major pathway to detoxify irinotecan,16 whereas the subsequent hepatobiliary excretion of irinotecan and its metabolites (SN-38 and SN-38G)16, 17 is mediated by transmembrane drug transporters such as ABCC2, ABCG2 and ABCB1.5, 12, 18

The prospective study conducted by Innocenti et al.19 has established the association of the UGT1A1*28 variant with severe toxicity and laid the foundation for genotype-based irinotecan dosing strategy. Many studies since then have revealed that other genetic polymorphisms in the UGT1A1 gene or in other genes that are involved in the metabolism and disposition pathways of irinotecan also predisposed individual subjects to the risk of severe adverse drug reactions (ADRs).11, 20, 21, 22, 23

In this study, we genotyped and comprehensively analyzed 170 single nucleotide polymorphisms (SNPs) in 14 genes that possibly have essential roles in the metabolism and transport of irinotecan 24, 25, 26 to verify SNPs associated with severe myelosuppression in the Japanese population (Figure 1). We then fine-mapped the genomic region containing the associated SNP, performed haplotype analysis and elucidated its association with severe myelosuppression in the Japanese population.

Figure 1
figure 1

Disposition and metabolic pathways of irinotecan (CPT-11).

Materials and methods

Subjects

This study was a retrospective case–control association study performed in two stages.27 In both stages, the cases refer to patients who developed severe myelosuppression (such as leukopenia, neutropenia and/or anemia of grade 3 or 4) during irinotecan therapy, whereas controls refer to subjects who did not show any sign of ADR during irinotecan therapy. Grade of ADR was classified according to the National Cancer Institute—Common Toxicity Criteria version 2.0.28

The first stage involved 344 subjects who received irinotecan therapy and who were registered into ‘the Leading Project for Personalized Medicine,’ or the BioBank Japan Project, in the Ministry of Education, Culture, Sports, Science and Technology, Japan, from June 2003 to December 2007.29 The BioBank Japan Project started in 2003 with the goal of collecting DNA and serum samples, along with clinical information, from 300 000 cases diagnosed with any of the 47 different diseases from a collaborative network of 66 hospitals in Japan. The biological materials and clinical information were collected from patients with written informed consent by medical coordinators at participating institutes.

Among the 344 aforementioned subjects, 23 and 58 subjects were each classified as cases and controls, whereas the remaining subjects developed either myelosuppression of lower grades or other ADRs such as nausea, vomiting, anorexia and/or diarrhea.

The second stage involved 63 subjects (39 subjects from Tokushukai Hospital Groups and 24 subjects from Sapporo Medical University) who received irinotecan therapy. Among these, 7 and 20 subjects were categorized as cases and controls, respectively. Table 1 describes the demographic characteristics of patients who participated in this study.

Table 1 Demographical characteristic of patients

This project was approved by the ethics committees at The Institute of Medical Science, The University of Tokyo, Tokushukai Hospital Groups and Sapporo Medical University, Japan.

Selection of SNPs and design of the study

For each individual who participated in the first stage, a total of 170 SNPs (tagSNPs (tSNPs) or cSNPs) in or near to the 14 candidate genes involved in the metabolism and transporting pathways of irinotecan (ABCB1, ABCC1, ABCC2, ABCC5, ABCG2, BCHE, CES2, CYP3A4, CYP3A5, SLCO1B1, SLCO1B3, UGT1A1, UGT1A7 and UGT1A9) were genotyped. As a whole, 27, 45, 12, 21, 8, 4, 9, 2, 1, 15, 18, 3, 4 and 1 SNPs were genotyped for each of the listed genes, respectively. All the tSNPs investigated in this study have reported minor allele frequencies of greater than 10% in the Japanese HapMap database (http://www.hapmap.org/) and capture most of the haplotypes in linkage disequilibrium (LD) blocks encompassing each gene (r2>0.8). However, lacking information on tSNPs, only several functional SNPs were tested for CYP3A4, CYP3A5, UGT1A1, UGT1A7 and UGT1A9. All SNPs were genotyped by using either the multiplex polymerase chain reaction-based Invader assay30 or by direct sequencing as were described in Cha et al.31

Statistical analyses

The genotype and allele frequencies of each SNP were calculated and tested with the standard χ2-test of the Hardy–Weinberg equilibrium (HWE).32 SNPs that showed a deviation from the HWE (Pχ2<0.05) in the control sample were eliminated from the subsequent analyses. Associations of each SNP with severe myelosuppression were evaluated by using Fisher's exact tests that consider each of the allelic, dominant-inheritance and recessive-inheritance models. SNPs showing a PFisher value of 0.01 or smaller in the first stage were considered as candidate SNPs that might be associated with severe myelosuppression and were further investigated in the second stage. As the number of subjects in the replication study is small, to increase the power of the study, genotyping results from both stages were combined and analyzed jointly. Bonferroni correction was applied for the judgment of statistical significance of the combined analysis.

Fine-mapping and haplotype analysis

According to the HapMap Japanese population database, other SNPs, which are in r2>0.3 with the SNP showing a significant association with severe myelosuppression in this study, were further genotyped by the Invader assay. In addition, haplotype analysis for the fine-mapped region was also performed by Haploview software v4.1.33

Results

Patient characteristics

A total of 108 patients were enrolled (81 in the first stage and 27 in the second stage) in this study. Nearly half of these subjects suffered from colorectal cancer (47 and 59%, respectively). The remaining patients suffered from lung cancer (27 and 22%, respectively), gastric cancer and ovarian cancer. Overall distributions of gender (male/female) were 18/12 in cases and 51/27 in controls (P=0.658). There was no distinctive difference in age distribution (median, years (range)) between the ADR and non-ADR groups (61.63 (37–81) versus 62.21 (34–84)). The patients with or without concomitant anticancer drugs were 2/28 or 13/65 in cases/controls, suggesting no significant difference (P=0.227). No significant difference was observed in the incidence of severe myelosuppression between patients of the first and second stages (23/58 versus 7/20) (P=1.00).

Association study of SNPs with irinotecan-induced severe myelosuppression

In the first stage, 23 patients who developed severe myelosuppression and 58 patients who did not show any signs of ADR after receiving irinotecan therapy were genotyped for a total of 170 SNPs in or near to the 14 genes that have critical roles in the metabolism and transport of irinotecan. We then examined associations of these SNPs with severe myelosuppression. Detailed information of the 170 SNPs and the results of association analyses with severe myelosuppression in the first stage of the study were summarized in Table 2.

Table 2 Associations of 170 SNPs with severe myelosuppression in subjects who received irinotecan therapy

In the process of the quality control of genotyping data, we excluded seven SNPs the genotype frequencies of which showed significant deviation from the HWE (Pχ2<0.05) in the control samples for further analyses. In addition, five markers that were found to be non-polymorphic in our samples of the first stage were excluded from subsequent analyses. Furthermore, we excluded eight SNPs (two in UGT1A7, two in ABCB1, one in ABCC2, two in SLCO1B3 and one in ABCC1) that were absolutely linked to another SNP in our study (r2=1, as were determined by Haploview 4.1 software33) from the subsequent analyses (Table 2).

Among the remaining 150 SNPs, ten showed minimum PFisher value of 0.05–0.01; these SNPs are located in the ABCB1 (three SNPs), ABCC1 (four SNPs), UGT1A1 (one SNP) and SLCO1B3 (two SNPs) (Table 2). In addition, five SNPs, two in the SLCO1B3 and one in each of the ABCC1, UGT1A7 and ABCG2, showed minimum PFisher value of less than 0.01. We considered these five SNPs as possible candidates and further genotyped them by using DNA samples of the 27 subjects who participated in the second stage of the study. These include 7 patients who developed severe myelosuppression and 20 who did not show any signs of ADR by the irinotecan therapy. Among the five SNPs examined, only one showed PFisher<0.05 in the second stage of the study. This SNP, rs2622604, is located in intron 1 of the ABCG2 gene (PFisher=0.0419) (Table 3).

Table 3 Results of second and combined studies for SNPs associated with severe myelosuppression with Fisher test's P<0.01 in the first study

Besides, when Bonferroni correction for multiple testing (based on 150 independent tests) was applied for judgment of statistical significance (α<0.000333) in the combination analysis, the SNP rs2622604 was the only one showing a significant level of association with severe myelosuppression, with a Bonferroni-corrected P-value of smaller than 0.05 (PFisher=0.000237, PCorrected=0.036). In addition to this SNP, rs7977213 in SLCO1B3 as well as the UGT1A7*3 variant revealed a relatively small P-value in the combination analysis, although it did not reach a significant level after Bonferroni correction.

Fine mapping of the ABCG2 gene and haplotype analysis

We further genotyped six other SNPs in the ABCG2 gene that, according to the HapMap database of the Japanese population, are in r2>0.3 with rs2622604. We identified another SNP, rs3109823, which shows a stronger association with severe myelosuppression (PFisher=0.0000133; Pχ2=3.46E-06) (Tables 4 and 5a). Haplotype analysis that considered the seven SNPs of the ABCG2 gene revealed that two haplotypes, consisting of the two aforementioned SNPs, showed an association as strong as the independent effects of the two SNPs (haplotype TC, Pχ2=2.49E-06; haplotype CT, Pχ2=6.71E-05) (Table 5b) (Figure 2).

Table 4 Fine-mapping of the ABCG2 gene
Table 5a Frequencies of associated allele and corresponding P-values for SNPs examined for ABCG2
Table 5b Frequencies of haplotypes and corresponding P-values for haplotypes of ABCG2
Figure 2
figure 2

Haplotype structure of the ABCG2 gene. Pairwise LD is displayed in black (strong LD) and gray (low LD) rectangles. Values represent r2 measurements.

Discussion

This study is a case–control association study in a retrospective design. Although the pharmacokinetic data for each subject could not be obtained, correlations between severe myelosuppression with 170 loci had been directly examined. In addition, statistical support that an intronic SNP of the ABCG2 gene, rs2622604, was likely to be associated with irinotecan-induced severe myelosuppression (grade 3 or 4) has been shown. On top of that, we have also fine-mapped the ABCG2 gene and identified another SNP, rs3109823, which showed stronger association with severe myelosuppression.

ABCG2 (ATP-binding cassette, Subfamily G, Member 2) encodes a transmembrane protein that mediates the hepatobiliary excretion of SN-38 and may have a major role in the pharmacokinetics of irinotecan.34 In addition to that, several studies have also reported that the overexpression of ABCG2 is associated with a decrease in the intracellular concentration of SN-38, leading to making cancer cells SN-38/irinotecan-resistant.35, 36, 37, 38 In this study, two SNPs in ABCG2 have been associated with severe myelosuppression that might be associated with an elevated level of SN-38 in systemic circulation. Furthermore, haplotypes containing the two SNPs have also shown a similarly strong association with severe myelosuppression. Although downstream functional analyses and pharmacokinetic studies were not performed to provide additional supporting evidences on the functional relevance of the two SNPs for severe myelosuppression and we have not pinpointed the causative SNP, a recent study by Poonkuzhali et al.39 has provided evidence that the SNP, rs2622604, was associated with a lower mRNA expression of ABCG2. Their results support our hypothesis that individuals who carry the risk genotypes for rs2622604 might have reduced hepatobiliary efflux activity of SN-38, leading to elevated intracellular concentration of SN-38 in hepatocytes. This subsequently causes the accumulation of irinotecan/SN-38 in the systemic circulation and induces severe myelosuppression. On the other hand, although rs3109823 revealed stronger association with severe myelosuppression in our study, this SNP did not seem to be associated with reduced mRNA expression according to the study of Poonkuzhali et al.39 The strong association of this SNP with severe myelosuppression might simply be because of its strong LD with rs2622604 (r2=0.70). Further functional analysis on these SNPs in the ABCG2 gene may provide additional evidence to support our hypothesis.

In addition to the ABCG2 gene, we also investigated associations of UGT1A variants such as UGT1A1*28, UGT1A1*6, UGT1A1*27 and UGT1A7*3 with severe myelosuppression in the exploratory study. We did not observe strong associations of UGT1A1*28 and UGT1A1*27 variants with severe myelosuppression possibly because of the low allelic frequencies of these variants in the subjects we examined. Although the UGT1A1*6 variant, which is more prevalent in the Asian population, revealed a weak association (PFisher=0.0214) (Table 2) with severe myelosuppression, we did not further investigate this variant because we observed much stronger association of the UGT1A7*3 variant than the UGT1A1*6 variant. In our study, rs11692021 that represents the *3 variant of the UGT1A7, which was reported to have the highest glucuronidation rate of SN-38 among UGT1A isoforms in vitro,40, 41 showed the strongest association with severe myelosuppression (PFisher=0.00583). Although none of the UGT1A variants showed significant association with severe myelosuppression in our study, the tendency of association could be observed. Thus, examining the association of these variants with severe myelosuppression in a larger number of subjects should be rewarding. As these variants are in strong LD,42 we need to further analyze which UGT1A variant(s) has biological significance.

Although the number of subjects investigated in this study is limited and the associations of several variants, such as those in the UGT1A genes, need to be further validated by examining a larger number of subjects, we have successfully identified an intronic SNP in the ABCG2 gene to be significantly associated with irinotecan-induced severe myelosuppression. Our results suggest that genetic polymorphism in ABCG2 might alter transport activity for the drug and elevate the level of irinotecan in systemic circulation, leading to severe myelosuppression.